Hi,
Having a lot of 'ipmitool sol activate' instances running on our
conserver(s), I have experienced clear and reproduceable memory leaks, which
eventually kill the server. We worked around this by cronkilling greedy ipmitool
processes.
The facts are these:
Normal ipmitool processes use around 2M of memory. Now some of these start
leaking relatively slowly (a few hundred kB a week). When trying to attach
to these consoles, they seem dead, and I need to reopen them (down/up the
console in conserver, which does a kill of the ipmitool process).
I did an strace on a normal and a leaking process, and there seems to be a
difference (Timeout):
## [Leaking process example]
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
sendto(3, "\6\0\377\7\6\300\0\10\0\2U,\0\0 \0{\7\220\270\376s\372"..., 64, 0,
NULL, 0) = 64
nanosleep({0, 100000}, NULL) = 0
select(4, [3], NULL, [3], {1, 0}) = 0 (Timeout)
nanosleep({0, 5000000}, NULL) = 0
sendto(3, "\6\0\377\7\6\300\0\10\0\2V,\0\0 \0\361\246S\235Z\217\6"..., 64, 0,
NULL, 0) = 64
nanosleep({0, 100000}, NULL) = 0
select(4, [3], NULL, [3], {2, 0}) = 0 (Timeout)
nanosleep({0, 5000000}, NULL) = 0
sendto(3, "\6\0\377\7\6\300\0\10\0\2W,\0\0 \0\332\252\210\236S\340"..., 64, 0,
NULL, 0) = 64
nanosleep({0, 100000}, NULL) = 0
select(4, [3], NULL, [3], {3, 0}) = 0 (Timeout)
nanosleep({0, 5000000}, NULL) = 0
sendto(3, "\6\0\377\7\6\300\0\10\0\2X,\0\0 \0\203\275}$\3509\314\350"..., 64,
0, NULL, 0) = 64
nanosleep({0, 100000}, NULL) = 0
select(4, [3], NULL, [3], {4, 0}) = 0 (Timeout)
nanosleep({0, 5000000}, NULL) = 0
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
## [Normal process example]
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
sendto(3, "\6\0\377\7\6\300\0\10\0\2\6\26\0\0 \0\233\333~\333e\333"..., 64,
0, NULL, 0) = 64
nanosleep({0, 100000}, NULL) = 0
select(4, [3], NULL, [3], {1, 0}) = 1 (in [3], left {0, 899000})
recvfrom(3, "\6\0\377\7\6\300\244\243\242\240\4\26\0\0000\0\363\252"...,
1024, 0, NULL, NULL) = 80
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
select(4, [0 3], NULL, NULL, {0, 500000}) = 0 (Timeout)
I tried several ipmitool versions from 1.8.8 to 1.8.11, recompiled from
source on a rhel3,4 and 5 box.
It happens on different BMCs, e.g. Dell PE1950, IBM x3550, etc.
Any idea what could be happening here?
------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables
unlimited royalty-free distribution of the report engine
for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Ipmitool-devel mailing list
Ipmitool-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ipmitool-devel
